Skip to content

Conversation

@hedaoyuan
Copy link
Contributor

@hedaoyuan hedaoyuan commented Dec 26, 2017

Copy link
Contributor

@qingqing01 qingqing01 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Almost LGTM.

size_t colWidth = outputHeight * outputWidth;
// Max col matrix height 256, Max col matrix width 1024
size_t stepColHeight = std::min(colHeight, (size_t)256);
size_t stepColWidth = std::min(colWidth, (size_t)2048);
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

static_cast<size_t>(256)
static_cast<size_t>(2048)

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Done.

for (size_t filterWidth : {3, 7}) {
for (size_t stride : {1, 2}) {
for (size_t padding : {0, 1}) {
for (size_t dilation : {1, 3}) {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe the test case can be reduced to speed up the unit testing time.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Done.

(imRowIdx - paddingHeight) >= inputHeight ||
(imColIdx - paddingWidth) < 0 ||
(imColIdx - paddingWidth) >= inputWidth) {
colData[colh * colWidthSize + colw] = T(0);
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

static_cast<T>(0)

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Done.

outputWidth});

resizeBuffer<Device>(stepColHeight * stepColWidth * sizeof(real));
colData = reinterpret_cast<real*>(memory_->getBuf());
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I see that the memory_ is not released in every layer. Do you think the if at most 2M extra workspace is used in every layer, we do not need to release it?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Done. Add release the memory.

Copy link
Contributor

@qingqing01 qingqing01 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM.

@hedaoyuan hedaoyuan merged commit a096c58 into PaddlePaddle:develop Dec 29, 2017
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants